Texture-based urban detection using contourlet coefficient on remote sensing imagery

Abstract

This paper presents an original urban detection on remote sensing images, which is based on efficient textural features extracted from contourlet transform domain. Since cities and towns are consisted of various objects, which have great differences in scale, directivity and texture characteristics, it is adaptive for contourlet transform to detecting urbans result from its discrete-domain multi-resolution and multi-direction expansion using nonseparable filter banks. Therefore, construction of contourlet coefficients can capture the intrinsic geometrical structure that is the key in visual information. On the other hand, we employ robust texture features in contourlet domain for recognizing urban regions from other land type by LS-SVM classifier. The performance of the proposed approach was validated by experiments carried on a data set of large images consisting of heavily negative samples.